一种带奇异点检测和补偿的GPR在线软测量方法
发布时间:2019-03-17 07:23
【摘要】:针对软测量方法实际应用中查询样本可能出现奇异点这一问题,提出一种带奇异点检测和补偿的高斯过程回归(Gaussian process regression,GPR)在线软测量方法。该方法首先对训练样本利用高斯过程回归方法进行建模,得到软测量模型;然后对新来查询样本采用改进拉依达准则进行奇异点检测,当新来查询样本被确定为奇异点时,利用辅助模型进行修补,然后再利用软测量模型对修补后查询样本点进行预测;否则,直接对新来查询样本点使用软测量模型进行预测,此方法能够有效确保新来查询样本点的有效性。通过对实际硫回收过程的数据进行实验仿真,进一步验证了所提方法的有效性。
[Abstract]:In order to solve the problem that singular points may appear in query samples in the practical application of soft sensing method, an on-line soft sensing method based on Gao Si process regression (Gaussian process regression,GPR with singularity detection and compensation is proposed. Firstly, the training samples are modeled by Gao Si process regression method, and the soft sensor model is obtained. Then the new query sample is detected by the improved Lajda criterion. When the new query sample is determined as the singular point, the auxiliary model is used for patching, and then the soft sensor model is used to predict the patched sample point. Otherwise, the soft sensor model is used to predict the new query sample points directly. This method can effectively ensure the validity of the new query sample points. The effectiveness of the proposed method is further verified by simulating the data of the actual sulfur recovery process.
【作者单位】: 江南大学轻工过程先进控制教育部重点实验室;江南大学物联网工程学院;
【基金】:国家自然科学基金(21206053,21276111) 江苏省“六大人才高峰”计划资助项目(2013-DZXX-043) 江苏高校优势学科建设工程资助项目(PAPD)
【分类号】:O212.1
本文编号:2442055
[Abstract]:In order to solve the problem that singular points may appear in query samples in the practical application of soft sensing method, an on-line soft sensing method based on Gao Si process regression (Gaussian process regression,GPR with singularity detection and compensation is proposed. Firstly, the training samples are modeled by Gao Si process regression method, and the soft sensor model is obtained. Then the new query sample is detected by the improved Lajda criterion. When the new query sample is determined as the singular point, the auxiliary model is used for patching, and then the soft sensor model is used to predict the patched sample point. Otherwise, the soft sensor model is used to predict the new query sample points directly. This method can effectively ensure the validity of the new query sample points. The effectiveness of the proposed method is further verified by simulating the data of the actual sulfur recovery process.
【作者单位】: 江南大学轻工过程先进控制教育部重点实验室;江南大学物联网工程学院;
【基金】:国家自然科学基金(21206053,21276111) 江苏省“六大人才高峰”计划资助项目(2013-DZXX-043) 江苏高校优势学科建设工程资助项目(PAPD)
【分类号】:O212.1
【相似文献】
相关期刊论文 前10条
1 杨忠华;弦法在奇异点处的一个改进格式[J];高等学校计算数学学报;1990年02期
2 苏步青;射影空间曲的可表奇异点(英文)[J];中国数学学报;1951年01期
3 李彩荣;平面曲线上奇异点的性态[J];工科数学;1993年04期
4 柳朝阳,,赵永成;平面偏移曲线奇异点分类及处理[J];郑州大学学报(自然科学版);1996年01期
5 刘先志;用切向力法寻求绕铅垂轴旋转导轨上滑动质点的奇异点及其稳条件[J];应用数学和力学;1983年01期
6 裘肖庚;;变态二阶曲面与奇异点[J];绍兴师专学报;1992年05期
7 陆慰萍;;转轮曲线上的奇异点[J];青海师范大学学报(自然科学版);1988年02期
8 张宗燧,吴中发,鞠长胜;关于微扰论振幅的奇异性[J];物理学报;1965年08期
9 方德植;;关于具有五级奇异点的某种平面曲线对[J];厦门大学学报(自然科学版);1961年03期
10 方德植;关于具有(m,n)阶可表示奇up点的某种平面曲线对的研究[J];数学学报;1963年02期
本文编号:2442055
本文链接:https://www.wllwen.com/kejilunwen/yysx/2442055.html